joe-khawand / juggling-rl

Reinforcement Learning project for robotic juggling.

Juggling-RL a Reinforcement Learning project

Introduction

This project is inspired by the paper Kai Ploeger, Michael Lutter and Jan Peters. High Acceleration Reinforcement Learning for Real-World Juggling with Binary Rewards,Conference on Robot Learning (CoRL), 2020.

We implemented two different reinforcement learning approches in order to teach a mujoco robotic model how to juggle:

• The first approach was to use Direct Policy Search methods to optimise an expert policy we crafted
• The second was to use deep reinforcement learning to determine the optimal policy for juggling

How to run

To run the project first install the following python dependencies using pip or conda depending on your environment:

• numpy
• matplotlib
• mujoco
• mediapy
• gymnasium
• tensorflow
• os
• visualkeras

Then clone the repository on your machine.

You will find two jupyter notebooks inside the src directory:

• RL_DPS is our implementation using the Direct Policy Search Method
• RL is our implementation using deep deterministic policy gradient

Simply run all the cells to see results.

Paper

For more information about the implementation please refer to our paper in the docs folder.